Upgrade from 1.21.1 to 1.23.2 in Openshift

[rodsvilla]

I have read this section https://github.com/k8ssandra/cass-operator?tab=readme-ov-file#upgrade-instructions. Do i need to do this

kubectl -n cass-operator delete deployment.apps/cass-operator
kubectl -n cass-operator delete service/cassandradatacenter-webhook-service
kubectl patch crd cassandradatacenters.cassandra.datastax.com -p '{"spec":{"preserveUnknownFields":false}}'

if my goal is to upgrade my operator from 1.21.1 to 1.23.2 in Openshift? Below is the screenshot of my current version

Below should be the version based on the channel

[burmanm]

No, that applies to 1.7.0 -> 1.8.1 update only.

[rodsvilla]

Would you have an idea why is it not proceeding to 1.23.2?

[burmanm]

What does it complain?

[rodsvilla]

Nothing really. But upon further digging, AI helped me find the result below:

`
This PackageManifest output is the "smoking gun." It explains exactly why OLM is refusing to upgrade you, and it confirms that your Catalog is actually up to date (it sees v1.23.2).

The Problem: The "Bridge" is Missing
Look at the entries section at the bottom of your output:

YAML

channels:

• currentCSV: cass-operator.v1.23.2
  entries:
  • name: cass-operator.v1.23.2
    version: 1.23.2
    Your stable channel only contains one version: v1.23.2.

OLM requires a connected "Upgrade Graph" to move from one version to another. It needs to see a path like: v1.21.1 -> v1.22.x -> v1.23.2.

Because v1.21.1 (your installed version) is missing from this list, OLM considers your current installation to be "orphaned." It doesn't know how to get from your current version to the new one because the history connecting them has been removed from the catalog. It reports AtLatestKnown because, as far as it can tell, your version doesn't exist in the channel at all, so it gives up trying to calculate a path.
`

[burmanm]

We don't push catalogs to Red Hat, so this would be generated from their side. Also, we haven't set limitations to upgrade paths, so I'm guessing this is something that got autogenerated between Openshift version upgrades or so. Maybe there's a force install possibility?

[burmanm]

My Openshift installations do not sadly have any with 1.21.1 already installed, so I only see 1.23.2 as an option in the OperatorHub (so I'm assuming there's some filtering we didn't do that's happening). Oddly enough, the community versions shows older versions also.

[rodsvilla]

Yes, reinstallation is my only option now. I already tried in my sandbox and it worked (got 1.23). Just a quick sanity check question though, if I deleted the operator's sub and csv, will my cassandra clusters act up?

Also, you mentioned that you do not control what is being pushed in the redhat-certified catalog. I have an ongoing Red Hat ticket with me and I'll ask them why is it that upgrade paths are not available.

[burmanm]

I recommend switching to Helm installations, I sadly don't know how OLM behaves in this case. As long as it doesn't remove the CRDs you should be fine, but too much of OLM is "magic" (and OLM itself is being redesigned so the current version is dead end).